Robust dynamic traffic assignment for single destination networks under demand and capacity uncertainty

In this article, we discuss the system-optimum dynamic traffic assignment (SO-DTA) problem in the presence of time-dependent uncertainties on both traffic demands and road link capacities. Building on an earlier formulation of the problem based on the cell transmission model, the SO-DTA problem is robustly solved, in a probabilistic sense, within the framework of random convex programs (RCPs). Different from traditional robust optimization schemes, which find a solution that is valid for all the values of the uncertain parameters, in the RCP approach we use a fixed number of random realizations of the uncertainty, and we are able to guarantee a priori a desired upper bound on the probability that a new, unseen realization of the uncertainty would make the computed solution unfeasible. The particular problem structure and the introduction of an effective domination criterion for discarding a large number of generated samples enables the computation of a robust solution for medium- to large-scale networks, with low desired violation probability, with a moderate computational effort. The proposed approach is quite general and applicable to any problem that can be formulated through a linear programing model, where the stochastic parameters appear in the constraint constant terms only. Simulation results corroborate the effectiveness of our approach.